52

A new method is introduced that applies the ‘etiologic fraction under the exposed’ calculation for a specific disease to mortality data; i.e. the mortality fraction MFe. The MFe can be further partitioned in a specific (SMF) and a general (GMF) component. An application is given for smoking and bladder cancer (BC) mortality using data from the Hong Kong mortality case-control study which includes all adult deaths in Hong Kong in 1998.

Multinomial logistic regression is used to calculate three odds ratios (OR); 1) 25,981 deaths (excluding BC deaths) are compared with 12,371 living persons to ascertain whether smoking is associated with death in general (ORdl); 2) 169 BC deaths are compared with all other deaths to ascertain whether smoking is specifically associated with BC death (ORcd); 3) The BC deaths are compared with the living persons (ORcl), this association contains both the general and the specific effect of smoking on BC death. Assuming that ORcd > 1, ORcl > ORcd and ORcl/ORcd = ORdl, we show that the SMF can subsequently be calculated as 1- (1/ORcd) and the GMF as 1 - (1/ORcl) - (1-(1/ORcd)). The SMF indicates the fraction of BC deaths in the exposed group that is directly related to the exposure. A high SMF may give clues about etiology, prognosis, and treatment options. A high GMF may be important from a public health perspective as it indicates that the risk factor exerts its effect indirectly by its general effect on death. Reduction of exposure to this factor reduces mortality in general and consequently BC mortality. The ORdl was 1.45 (95% CI 1.31, 1.61) for current smokers and 1.21 (95% CI 1.12, 1.29) for ex-smokers, ORcl was respectively 3.18 (95% CI 1.82, 5.56) and 1.78 (95% CI 1.20, 2.65) and the ORcd 2.19 (95% CI 1.25, 3.84) and 1.48 (95% CI 1.00, 2.20). The SMF for BC mortality was 0.54 (95% CI 0.19, 0.90) for current smokers and 0.32 (95% CI 0.08, 0.57) for ex-smokers indicating a stronger specific effect in current smokers. The GMF for current smokers was 0.14 (95% CI 0.02, 0.27). Although the ORs show a reduced risk for ex-smokers, the SMF and GMF (0.12, 95% CI 0.05, 0.17) for ex-smokers indicate that among ex-smoking BC deaths still a large proportion of the mortality can be explained by their previous smoking habits. Smoking remains therefore an important risk factor for BC mortality in ex-smokers.

98th AACR Annual Meeting-- Apr 14-18, 2007; Los Angeles, CA